Stochastic based extended krylov subspace method for power/ground network analysis

Abstract

In this paper, we present a novel stochastic simulation approach based on extended Krylov subspace method for on-chip power grid analysis. The new method performs the analysis by using random walk in a stochastic manner. But different from the existing random walk method, the moments of the circuits are computed and extended Krylov subspace (EKS) method is used to calculate the responses in frequency domain. The new method can compute the transient responses in a local manner, which is in contrast to the existing random walk method, thus improves the existing frequency-domain random walk method by using extended Krylov subspace method. The resulting method is more numerically stable and faster than existing random walk methods. Experimental results demonstrate the advantages of the proposed method, called rwEKS, over EKS method for localized power grid analysis.